Appl Water Sci (2016) 6:371–381 DOI 10.1007/s13201-014-0230-6

ORIGINAL ARTICLE

Assessing groundwater quality for irrigation using indicator kriging method

Masoomeh Delbari • Meysam Amiri • Masoud Bahraini Motlagh

Received: 18 January 2013 / Accepted: 1 September 2014 / Published online: 14 September 2014 Ó The Author(s) 2014. This article is published with open access at Springerlink.com

Abstract One of the key parameters influencing sprinkler probability maps. The suitable areas for sprinkler irrigation irrigation performance is water quality. In this study, the design were determined to be 25,240 hectares, which is spatial variability of groundwater quality parameters (EC, about 34 percent of total agricultural lands and are located ? - - SAR, Na ,Cl, HCO3 and pH) was investigated by in northern and eastern parts. Overall the results of this geostatistical methods and the most suitable areas for study showed that IK is an appropriate approach for risk implementation of sprinkler irrigation systems in terms of assessment of groundwater pollution, which is useful for a water quality are determined. The study was performed in proper groundwater resources management. Fasa county of using 91 water samples. Results indicated that all parameters are moderately to Keywords Groundwater quality Á Geostatistics Á GIS Á strongly spatially correlated over the study area. The spa- Probability map Á Sprinkler irrigation - tial distribution of pH and HCO3 was mapped using ordinary kriging. The probability of concentrations of EC, SAR, Na? and Cl- exceeding a threshold limit in Introduction groundwater was obtained using indicator kriging (IK). The experimental indicator semivariograms were often is located in arid and semi-arid part of the world and fitted well by a spherical model for SAR, EC, Na? and Cl-. its water resources are in a critical condition. In Iran, the - For HCO3 and pH, an exponential model was fitted to the average annual rainfall is less than 250 mm, which is less experimental semivariograms. Probability maps showed than one-third of the average rainfall in the world that the risk of EC, SAR, Na? and Cl- exceeding the given (800 mm). The average annual evaporation is about critical threshold is higher in lower half of the study area. 2,100 mm and it is almost three times the world’s average The most proper agricultural lands for sprinkler irrigation annual evaporation (700 mm). Iran’s population is implementation were identified by evaluating all approximately one percent of the world population, while its share of total renewable water resources of the world is only 0.36 percent. Moreover, the spatial and temporal & M. Delbari ( ) distribution of rainfall is not uniform; 70 % of rainfall is Department of Water Engineering, Faculty of Water and Soil, University of Zabol, Zabol, Iran distributed over 25 % of the country (Alizadeh 2004). In e-mail: [email protected] addition, most of the rainfall occurs in non-irrigation sea- sons. So, groundwater resources play a key role in irriga- M. Amiri tion of arid and semi-arid agricultural lands. However, the Scientific Staff in the Department of Water Resources, International Hamoun Wetland Research Institute, University of suitability of groundwater quality for irrigation purposes Zabol, Zabol, Iran requires to be considered in the beginning for a sustainable utilization of groundwater. Groundwater quality could be M. B. Motlagh influenced by different sources of pollution such as Graduated Student of Water Resources Engineering, Department of Water Engineering, Faculty of Water and Soil, University of domestic waste water, urban runoff, intensive application Zabol, Zabol, Iran of fertilizers and solid waste disposal (Hammouri and El- 123 372 Appl Water Sci (2016) 6:371–381

Naqa 2008). On the other hand, most agricultural areas are which was deemed to be more easily available at larger irrigated by traditional surface irrigation methods while sample size. Indicator kriging is a non-parametric geosta- their irrigation efficiency never exceeds 35 % (Alizadeh tistical estimation method, which has the ability to produce 2004). To improve irrigation efficiency and to increase probability maps in which the probability of any soil and water use efficiency in agriculture, a considerable budget water pollutant being larger or smaller than a given thresh- has been allocated for agricultural sector to adapt modern old value is provided (Goovaerts 1997; Cullmann and Sab- irrigation systems effective from 2010. Often, the farmers orowski 2005). Indicator kriging has been used frequently pay only 15 % of the costs for implementation of sprinkler for studying the risk of soil or groundwater contamination irrigation system, and the remaining 85 % of costs is paid (Juang and Lee 2000; van Meirvenne and Goovaerts 2001; by the government. With the modern irrigation methods Goovaerts et al. 2005;Jangetal.2010). Indicator kriging however, the water quality should be carefully controlled to was used by Liu et al. (2004) to assess arsenic contamination prevent water applicator clogging or water distribution in an aquifer in Yun-Lin, Taiwan. Goovaerts et al. (2005) equipments’ damage. In addition, utilization of polluted studied the spatial variability of arsenic in groundwater in irrigation water deteriorates soil physical properties and southeast Michigan using indicator kriging. Dash et al. reduces the crop yield (Shainberg and Oster 1978; Solomon (2010) used indicator kriging to assess the probability of 1985; Lauchli and Epstein 1990; Shahalam et al. 1998). exceedence of groundwater quality parameters e.g. EC, Cl- - Therefore to achieve a more effective conservation of and NO3 in Delhi, India. Ordinary and indicator kriging natural resources, agricultural management of soils and were used by Adhikary et al. (2010) for mapping ground- water becomes essential (Delgado et al. 2010). water pollution in India. They evaluated the spatial vari- 2? - - ? 2- 2? The main water characteristics which determine the ability of Ca ,Cl ,HCO3 ,EC,Na ,SO4 ,Mg and - suitability of ground water for irrigation according to NO3 and produced pollution risk maps for these parame- United States Salinity Laboratory (USSL 1954) include: (1) ters. In another study, Adhikary et al. (2011) used indicator total concentration of soluble salts; (2) relative ratio of and probability kriging methods for delineating Cu, Fe, and sodium to other cations (Mg2?,Ca2? and K?) (3) con- Mn contamination in groundwater in Delhi, India. Piccini centration of boron, chloride or other toxic elements and et al. (2012) illustrated an application of indicator kriging for (4) under some conditions, carbonate or bicarbonate con- mapping the probability of exceeding nitrate contamination centrations that is associated with Ca2? and Mg2? con- thresholds in Central Italy. centrations. Both physical and chemical qualities of water The aim of this study was to evaluate the potential of the are important in irrigation systems performance. With groundwater to cause crop problems through toxicity, sprinkler irrigation in which water sprayed over the plants, salinity and soil infiltration rate for determining the suit- low quality of water may cause severe damage, such as leaf ability of groundwater for sprinkler irrigation uses. For this, burn (Keller and Bliesner 1990). Most trees are sensitive to the spatial variability of groundwater quality parameters in even relatively low concentrations of sodium and chloride. Fasa county of Fars province was investigated using geo- Increase in bicarbonate concentration of irrigation water in statistical tools. Ordinary and indicator kriging were used sprinkler irrigation causes white spots on leaves or fruits. to map the spatial distribution and the probability of An increase in calcium carbonate of irrigation water in drip exceeding a critical threshold for some water contaminants. irrigation causes precipitation of particles and emitter clogging (Shatnawi and Fayyad 1996). Point data are usually inadequate to identify vulnerable Materials and methods areas, so appropriate interpolation methods are needed to map the spatial distribution pattern of groundwater quality Description of study area parameters. In contract to traditional interpolation methods which are unable to map uncertainty of the predicted values, Fasa county in southern Iran lies between 53° 190 and 54° geostatistical-based algorithms can be used to assess the 150 east longitude and 28° 310 and 29° 240 north latitude uncertainty attached to the estimated values for incorporat- (Fig. 1). This county is surrounded by the counties of ing in decision making (Van Meirvenne and Goovaerts and from the north, by the counties of 2001). Several studies have been done to investigate spatial and Zarindasht from the south, by the counties of variability of groundwater quality using geostatistics. Coo- and Estahban from the east and by the counties of per and Istok (1988) utilized kriging for estimation of Shiraz and Jahrom from the west. The 30-year average groundwater contaminants at the Chem-Dyne Superfund site rainfall is 285.6 mm, the average air temperature is 18 °C in Ohio. Istok et al. (1993) used cokriging to estimate her- and the average annual evaporation is 2,610 mm. The bicide concentration of groundwater in an agricultural elevation from the sea level generally decreases from the region. The auxiliary variable was nitrate concentration north to the south; the central and southern parts of the 123 Appl Water Sci (2016) 6:371–381 373

Fig. 1 Location map of study area and observation wells region have the lowest elevation, while the northern parts are surrounded by mountains. The highest and lowest ele- vations over the study area are 3,172 and 1,117 m from free sea level, respectively. In recent years, due to successive droughts, the govern- ment’s macro policies in agricultural sector have focused on the implementation of modern irrigation systems. Each year, large public subsidies are allocated to this sector. Sprinkler irrigation systems have gained more attention in the study area from 2000 to 2010. From 2007, however, the imple- mentation rate of sprinkler irrigation has increased with the greater slope, which may be due to higher subsidies (50 %). From 2010 onwards, the farmers pay only 15 % of the cost for implementation of sprinkler irrigation systems, and the remaining 85 % of the cost is paid by the government; so sprinkler irrigation systems are expected to be more wel- come. Based on the land use map (Fig. 2), the study area consists of farm lands, forest, pasture and dry lands. The sprinkler irrigation systems are supposed to be implemented in agricultural lands (dry and irrigated lands), which cover about 74,032 hectares of the whole area. Fig. 2 Land use map of the study area

123 374 Appl Water Sci (2016) 6:371–381

Groundwater sampling and analysis where Z(xi) and Z(xi ? h) are the measured values at locations x and x ? h, respectively. N(h) is the number of In this study, groundwater quality data from 91 observation i i pairs of data separated by vector h. After calculating wells obtained in 2011 were used for the analysis. The experimental semivariogram, a theoretical model should be location of the observation wells are shown in Fig. 1. fit to the experimental data. The type of model and its Groundwater samples were analyzed for chemical param- characteristics are used in kriging system to interpolate the eters including pH, Na?,Cl-, HCO -, electrical conduc- 3 desired variable. The most common types of semivario- tivity (EC) and sodium absorption ratio (SAR) by adapting gram models are spherical, exponential and Gaussian the standard procedures of water analysis. (Isaaks and Srivastava 1989). The evaluation of water quality is a fundamental To investigate the spatial variability of groundwater requirement for implementation of modern irrigation sys- quality parameters, the experimental semivariograms of the tems. Richards (USSL 1954) proposed a table for the selected parameters are calculated in four directions 0, 45, classification of irrigation water based on EC and SAR. He 90 and 135° with an angle tolerance of 22.5°. also presented a diagram for the classification of irrigation water. Today there are many other ways for the classifi- Ordinary kriging cation of irrigation water but generally, the specified limit for all of these methods is nearly identical. One of the Ordinary kriging (OK) estimator so called ‘‘the best linear important water quality standards for the sprinkler irriga- unbiased estimator (BLUE)’’; is defined as follows (Journel tion systems that has been used in this study (Table 1)is and Huijbregts 1978): provided by Ayers and Westcot (1994). Xn Xn à Z ðx0Þ¼ ki Á ZðxiÞ with ki ¼ 1 ð2Þ Geostatistical analysis i¼1 i¼1 * A geostatistical analysis including the investigation of where Z (x0) is the estimated value at unsampled location spatial autocorrelation and the interpolation of attribute x0, ki is the weight assigned to the measured value Z(xi) and values at unsampled locations is performed in this study. n is the number of neighboring points. OK weights ki are Spatial autocorrelation analysis describes spatial continuity allocated to the known values in such that they sum to unity based on a experimental semivariogram. The experimental (unbiaseness constraint) and they minimize the kriging semivariogram, c*(h), can be defined as one-half the vari- estimation variance. The weights are determined by solving ance of the difference between the attribute values at all the following system of equations (Isaaks and Srivastava 1989): points separated by h as follows: (Isaaks and Srivastava 8 1989; Goovaerts 1997): > Xn > <> kjcðxi; xjÞþl ¼ cðxi; x0Þ; i ¼ 1; ...; n XNhðÞ 1 j¼1 cà h Zx Zx h 2 1 ð3Þ ðÞ¼ fgðÞÀi ðÞi þ ð Þ > Xn 2NhðÞi 1 > ¼ :> kj ¼ 1 j¼1

Table 1 Threshold limits for some irrigation water quality parameters (Ayers and Westcot 1994)

Irrigation problem Unit Degree of restriction on use

None Slight to moderate Severe

Salinity (EC) dS/m \0.7 0.7–3.0 [3 Na? meq/L \3 3–9 [9 Cl- meq/L \3 [3 Infiltration SAR= 0–3 And EC= [0.7 0.7–0.2 \0.2 3–6 [1.2 1.2–0.3 \0.3 6–12 [1.9 1.9–0.5 \0.5 12–20 [2.9 2.9–1.3 \1.3 20–40 [5.0 5.0–2.9 \2.9 - HCO3 meq/L \1.5 1.5–8.5 [8.5 pH – Normal range 6.5–8.4

EC electrical conductivity, SAR sodium absorption ratio

123 Appl Water Sci (2016) 6:371–381 375

where c(xi, xj) is the average semivariance between all pairs Semivariogram analysis of (binary) data and OK is of data locations; l is the Lagrange parameter for the performed using the software (GS ? (Gamma Design minimization of kriging variance and c(xi, x0) is the Software) 2006). IK is implemented using SGeMS (Remy average semivariance between the location to be estimated et al. 2009). The software package ArcGIS 9.0 (ESRI (x0) and the ith sample point. (Environmental Systems Research Institute Inc) 2004) with OK also provides a measure of uncertainty attached to Geostatistical Analyst Extensions is used for mapping. each estimated value through calculating the OK variance: Xn 2 r ðx0Þ¼ kicðxi; x0Þþl ð4Þ Results and discussion i¼1 This estimation variance may be used to generate a Statistical analysis confidence interval for the corresponding estimate - ? - assuming a normal distribution of errors (Goovaerts, 1997). The summary statistics of EC, SAR, Cl ,Na , HCO3 and pH and data are presented in Table 2. The maximum value - of pH and HCO3 is less than the irrigation water threshold Indicator kriging values given in Table 1 and from this perspective, there is no limitation in using groundwater for sprinkler irrigation Instead of real data, IK works with binary variable called over the study area. Comparing the mean and maximum ‘‘indicator variable’’. An indicator variable can only takes values of SAR, Na?, EC and Cl- with the threshold values the values of zero or one considering some threshold val- given in Table 1 indicate that most restrictions are related ues. For a continuous variable Z(xi), the indicator variable to the high level of SAR and Cl- and Na? concentrations I(xi;zk) where zk is the desired threshold limit is defined as in the groundwater. The frequency distribution of water follows (Goovaerts 1997): quality data (not shown) shows that 59 % of Cl-,57%of ( Na?, 26 % of SAR and 23 % of EC data values exceed 1ifZðxiÞzk their corresponding irrigation water threshold limits. The Iðxi; zkÞ¼ ; k ¼ 1; ...; K 0 otherwise maximum concentrations of Na? (48.5 meq/L), Cl- ð5Þ (92.5 meq/L) and EC (13.59 dS/m) are seen in the southern part of the study area with the lowest topographical ele- where K is the number of cutoffs. The experimental vation. High levels of SAR values are seen in south as well * indicator semivariogram, cI (h), is then defined for every set as north of the study region. of indicators at each cutoff zk as: ArcHydro tool in GIS is used to produce the waterways NðhÞ network by using a digital elevation model (DEM) of the 1 X cÃðhÞ¼ ½ŠIðx ; z ÞÀIðx þ h; z Þ 2 ð6Þ study area. It has been found that water moves from the I 2NðhÞ i k i k i¼1 north to the south so it can be said that use of pesticides and where N(h) is the number of pairs of indicator transforms chemical fertilizers on the upstream farms reduces the downstream water quality. These pollutants are transmitted I(xi; zk) and I(xi ? h; zk) separated by vector h. The conditional cumulative distribution function (ccdf) at through surface runoff and groundwater toward low ele- vated parts in south and decrease the quality of water there. location x0 is then obtained by the IK estimator as follows:

Xn Geostatistical analysis Fðx0; zkjðnÞÞ ¼ I Ãðx0; zkÞ¼ kiIðxi; zkÞð7Þ i ¼1 The results of semivariogram analysis show no consider- where I*(x0;zk) is the estimated indicator value at unsam- able anisotropy so the isotropic semivariogram is used for pled location x0 and ki is the weight assigned to the known further analysis. The experimental (indicator) semivario- indicator value I(xi;zk). These discrete probability functions gram of the quality parameters are computed and fitted must be interpolated within each class (between every two with the best theoretical semivariogram model. For EC, parts of ccdf) and extrapolated beyond the minimum and SAR, Na? and Cl-, 5 indicator cutoffs corresponding to maximum values to provide a continuous ccdf covering all 0.10, 0.25, 0.50, 0.75 and 0.90 quantiles of observed data the possible ranges of the property of interest (Goovaerts were selected. The properties of the best fitted model in 1997). Local uncertainty measures, e.g., conditional vari- terms of the lowest residual sum of squares (RSS) and the ance and probability of exceeding or not exceeding a given highest R2 has been presented in Table 3. Meanwhile a threshold, are provided through post processing of IK- cross-validation analysis is performed to confirm the based ccdf’s. semivariogram model types selected. The best indicator 123 376 Appl Water Sci (2016) 6:371–381

Table 2 Descriptive statistics of groundwater quality parameters Parameter Minimum Maximum Mean Standard deviation Skewness Kurtosis

EC (dS/m) 0.35 13.59 2.28 2.16 2.33 7.85 SAR 0.28 11.32 2.23 1.98 2.06 6.00 Cl- (meq/L) 0.45 92.50 9.11 13.04 3.74 19.24 Na? (meq/L) 0.40 48.50 7.42 9.18 2.55 7.95 - HCO3 (meq/L) 3.00 9.50 5.20 1.34 0.75 0.41 pH 6.85 8.50 7.39 0.31 0.78 0.91 EC electrical conductivity, SAR sodium absorption ratio

Table 3 Characteristics of the best model of (indicator) semivariograms for groundwater quality parameters

2 Groundwater quality parameter Quantile Model type C0 C0 ? CA0 (m) C/(C0 ? C) R RSS

EC (dS/m) 0.1 Spherical 0.0125 0.12 35,020 0.899 0.765 3.52E-03 0.25 Spherical 0.0903 0.23 38,990 0.612 0.845 3.45E-03 0.5 Spherical 0.115 0.30 49,400 0.615 0.933 2.26E-03 0.75 Spherical 0.0902 0.21 43,100 0.577 0.383 2.74E-03 0.9 Exponential 0.0144 0.09 8,760 0.847 0.884 2.29E-04 SAR 0.1 Spherical 0.0398 0.1 51,000 0.604 0.746 1.09E-03 0.25 Spherical 0.0747 0.22 34,800 0.656 0.841 3.22E-03 0.5 Spherical 0.143 0.29 49,200 0.512 0.931 1.54E-03 0.75 Exponential 0.038 0.17 6,720 0.775 0.948 2.52E-03 0.9 Spherical 0.019 0.09 12,510 0.781 0.736 1.36E-03 Cl- (meq/L) 0.1 Spherical 0.0001 0.13 38,090 0.999 0.727 6.8E-03 0.25 Spherical 0.098 0.23 92,870 0.77 0.86 2.9E-03 0.5 Spherical 0.12 0.32 55,500 0.625 0.934 2.55E-03 0.75 Spherical 0.0844 0.21 43,100 0.599 0.923 1.32E-03 0.9 Spherical 0.0001 0.08 6,510 0.999 0.785 9.50E-04 Na? (meq/L) 0.1 Spherical 0.046 0.09 43,000 0.489 0.564 2.43E-03 0.25 Spherical 0.101 0.201 42,000 0.498 0.809 4.49E-03 0.5 Spherical 0.091 0.28 45,000 0.675 0.915 6.5E-03 0.75 Spherical 0.069 0.20 63,680 0.655 0.904 6.83E-04 0.9 Exponential 0.001 0.091 8,730 0.989 0.476 1.59E-03 - HCO3 (meq/L) – Exponential 0.978 2.125 6,110 0.519 0.703 0.296 pH – Exponential 0.053 0.035 2,480 0.851 0.608 1.73E-4

2 C0 random variance, C structured variance, A0 range of influence, R correlation coefficient, RSS residual sum of squares, EC electrical conductivity, SAR sodium absorption ratio

? - - semivariogram model for EC, SAR, Na and Cl is often The semivariogram characteristics of pH and HCO3 - spherical and the best semivariogram model for HCO3 (Table 3) are used in OK system to map the spatial dis- and pH is exponential. Adhikary et al. (2010) suggested tribution of these parameters. The generated maps are a spherical semivariogram model for EC, Na?,Cl- shown in Fig. 3. According to the map of pH spatial dis- - and HCO3 . The spherical model was also selected for tribution, western half of the study area has slightly lower - ? - - EC, SAR, Cl ,Na and HCO3 by Delgado et al. pH than the eastern half. Figure 3 also shows that HCO3 (2010). values range from 4 to 7.5 (meq/L) over majority of the The higher the ratio of the structured variance to the study area. Small parts in the northeast appear to have - total variance, i.e., C/(C ? C0) is, the stronger is the spatial HCO3 values of 3–4 (meq/L). Considering the irrigation correlation. Accordingly, Na has the highest within-class water limits in Table 1 and Fig. 3 shows no limitation in - - spatial correlation while HCO3 has the lowest spatial using groundwater for irrigation in terms of pH and HCO3 continuity over the study area. The range of (within-class) values. Also shown in Fig. 3, the maps of OK-estimation spatial correlation is smallest for pH and highest for Cl-. standard deviation (sd), which measures the uncertainty

123 Appl Water Sci (2016) 6:371–381 377

- Fig. 3 Maps of spatial distribution and associated standard deviation for pH (up) and HCO3 (bottom) produced by OK

- associated with the pH and HCO3 . The estimation error IK is used to generate the probability map of exceeding (uncertainty) provided by OK is smaller at the sampling 3, 3 meq/L, 3 and 3 dS/m, respectively, for Cl-,Na?, SAR points and nearby locations and it becomes higher as the and EC (Fig. 4). distance between the observations is getting larger and For EC, the threshold limit of 3 dS/m is selected where no data are available (northern and eastern parts of because according to Table 1, for EC above 3 dS/m, there the study area). is severe restriction on use of water for irrigation while for

123 378 Appl Water Sci (2016) 6:371–381

Fig. 4 Probability maps of exceeding 3, 3 meq/L, 3 and 3 dS/m, respectively, for Cl-,Na?, SAR and EC (left) and classification of groundwater quality into suitable and unsuitable for sprinkler irrigation use based on the probability maps (right)

EC values below 3 dS/m, there is no (where EC is below about 90 % of study area, EC of groundwater is above 0.7 dS/m) and a slight to moderate restriction on use. SAR 0.7 dS/m and therefore there is no restriction on use. For of irrigation water except for a few locations is below 6, so SAR between 3 and 6, where EC is above 1.2 dS/m, there under this situation according to Ayers and Westcot (1994) is no restriction and where EC is between 0.3 and 1.2 dS/ if EC is below 0.3 dS/m, there is a severe restriction on m, there is a slight to moderate restriction on use (Ayers use. As EC of groundwater is above 0.5 dS/m, there is no and Westcot 1994). Therefore, we used a threshold limit of severe restriction on use. Also for SAR between 0 and 3, in 3 for SAR. For both Na? and Cl-, a critical limit of 3 meq/

123 Appl Water Sci (2016) 6:371–381 379

Fig. 4 continued

L is considered for producing probability maps. According higher in lower half of the study area. Based on probability to Ayers and Westcot (1994) there is a slight to moderate maps and using the probability threshold, which is assumed restriction on use of irrigation water with Na? and Cl- to be equal to the marginal probability of each parameter, above 3 meq/L for sprinkler irrigation. So the probability groundwater quality is classified as suitable and unsuitable maps of exceeding 3 dS/m, 3, 3 meq/L and 3 are produced for sprinkler irrigation purposes (Fig. 4). According to the for EC, Cl-,Na? and SAR, respectively. maps generated (Fig. 4), the unsuitable groundwater for As shown in Fig. 4, the probability of Cl-,Na?, SAR irrigation in terms of high amounts of Cl-,Na?, SAR and and EC exceeding their corresponding critical thresholds is EC are located in southern parts of the study area. These

123 380 Appl Water Sci (2016) 6:371–381

Conclusions

In this study, a geostatistical analysis of EC, SAR, Na?, - - Cl , HCO3 and pH of groundwater in Fars province, southern Iran is performed and suitability of groundwater for irrigation is assessed. Semivariogram analysis shows - that there is a moderate spatial correlation of HCO3 and pH and a moderate to strong within-class spatial correlation of EC, SAR, Na? and Cl- over the study area. The pH and - HCO3 values are found to be within the permissible limits of standard irrigation water quality. The generated maps of EC, SAR, Na? and Cl- show that water quality decreases from the north to the south of the study area, so the downstream lands have the most polluted groundwater. Thus, the use of groundwater for irrigation in these areas will damage crops and reduce yield. However, more tol- erable crops could be cultivated with a good drainage system installation to prevent soil salinization. The most suitable regions for the implementation of the sprinkler irrigation systems are located in northern and eastern parts. Large investments like sprinkler systems require careful consideration of options before implementation. The results of this study show that a combined utilization of Fig. 5 Classification of groundwater quality into suitable and geostatistics and GIS can be useful in decision-making unsuitable for sprinkler irrigation purpose across the agricultural lands processes such as identifying suitable areas for imple- mentation of the sprinkler irrigation systems and reducing the risk of losing the national capital. areas are supposed to be inappropriate for the implemen- tation of sprinkler irrigation. However, the suitability of Open Access This article is distributed under the terms of the irrigation water regarding salinity management should be Creative Commons Attribution License which permits any use, dis- tribution, and reproduction in any medium, provided the original assessed considering salt tolerance of the cultivated crops author(s) and the source are credited. and soil characteristics. This means in areas with slightly high levels of salinity and sodium, groundwater can still be used for irrigating certain crops e.g. barley, sugar beet and References wheat because different crops have different tolerance levels of salinity/sodium hazard (Bauder 2007). Also to Adhikary PP, Chandrasekharan H, Chakraborty D, Kamble K (2010) reduce leaf burn under sprinkler irrigation from both Assessment of groundwater pollution in West Delhi, India using sodium and chloride, irrigation on cool, cloudy days or geostatistical approach. Environ Monit Assess 167:599–615 Adhikary PP, Dash JCh, Bej R, Chandrasekharan H (2011) Indicator overnight would be beneficial. Moreover to avoid direct and probability kriging methods for delineating Cu, Fe, and Mn contact of saline water with leaf surfaces in sprinkler irri- contamination in groundwater of Najafgarh Block, Delhi, India. gation, drop nozzles and drag hoses are recommended Environ Monit Assess 176:663–676 (Bauder 2007). Alizadeh A (2004) Irrigation water quality. publications Astan Quds Razavi (In Persian) Considering the classification maps shown in Fig. 4,a Ayers RS, Westcot DW (1994) ‘‘Water quality for agriculture,’’Irrigation single map is produced to show location of suitable and and drainage Paper 29 Rev. 1 (Reprinted 1989, 1994). Available at unsuitable groundwater over the study area in terms of all http://www.fao.org/DOCREP/003/T0234E/T0234E00.htm considered groundwater quality parameters across the Bauder TA (2007) Colorado State University Extension water quality specialist; Waskom RM, Extension water resource specialist; agricultural lands identified in Fig. 2 (Fig. 5). According to and Davis JG., Extension soils specialist and professor, soil and this map, the most suitable areas for sprinkler irrigation crop science system design in terms of water quality are located in Cooper RM, Istok JD (1988) Geostatistics applied to groundwater north, northwest and east parts of the study area. The contamination. II: application. J Environ Eng, ASCE 114(2):287–299 appropriate areas for the implementation of the sprinkler Cullmann AD, Saborowski J (2005) Estimation of local probabilities irrigation systems are determined to be 25,240 hectares that for exceeding threshold values in environmental monitoring. Eur is about 34 % of total agricultural lands. J Forest Res 124:67–71

123 Appl Water Sci (2016) 6:371–381 381

Dash JP, Sarangi A, Singh DK (2010) Spatial Variability of Keller J, Bliesner RD (1990) Sprinkle and Trickle Irrigation. Van groundwater depth and quality parameters in the national capital Nostrand Reinhold, New York. ISBN 0-442-24645-5 territory of Delhi. Environ Manage 45:640–650 Lauchli A, Epstein E (1990) Plant responses to saline and sodic Delgado C, Pacheco J, Cabrera A, Batllori E, Orellana R, Bautista F conditions. Agricultural Salinity Assessment and Management (2010) Quality of groundwater for irrigation in tropical karst Manual, Tanji KK (ed.), ASCE, New York, 113–137 environment: the case of Yucatan, Mexico. Agr Water Manage Liu CW, Jang CS, Liao CM (2004) Evaluation of arsenic contam- 97:1423–1433 ination potential using indicator kriging in the Yun-Lin aquifer ESRI (Environmental Systems Research Institute Inc) (2004) ArcGIS (Taiwan). Sci Total Environ 321:173–188 9. Getting Started with ArcGIS. ESRI, Redlands Piccini C, Marchetti A, Farina R, Francaviglia R (2012) Application Goovaerts P (1997) Geostatistics for natural resources evaluation. of indicator kriging to evaluate the probability of exceeding Oxford University Press, New York nitrate contamination thresholds. Int J Environ Res 6(4):853–862 Goovaerts P, AvRuskin G, Meiliker J, Slotnick M, Jacquez G, Nriagu Remy N, Boucher A, Wu J (2009) Applied geostatistics with SGeMS: J (2005) Geostatistical modeling of the spatial variability of a users’ guide. Cambridge University Press, Cambridge arsenic in groundwater of southeast Michigan. Water Resour Res Shahalam A, Abu Zahra BM, Jaradat A (1998) Wastewater irrigation 41:1–19 effect on soil, crop and environment: a pilot scale at Irbid, GS ? (Gamma Design Software) (2006) Geostatistics for the envi- Jordan. Water Air Soil Pollut 106(3/4):425–455 ronmental sciences. GS ? User’s Guide. Version 7. Plainwell, Shainberg I, Oster JD (1978) Quality of irrigation water. IIIC MI, USA Publication No. 2, Volcani Center, P.O.Box 49, Bet Dagan, Hammouri N, El-Naqa A (2008) GIS based hydrogeological vulner- Israel, 65 pp. ISBN 92-9019-002-9 ability mapping of groundwater resources in Jerash area-Jordan. Shatnawi M, Fayyad M (1996) Effect of khirbet as-samra treated Geofisica Int 47:85–97 effluent on the quality of irrigation water in the central Jordan Isaaks E, Srivastava RM (1989) An introduction to applied geosta- Valley. Water Res 30:2915–2920 tistics. Oxford University Press, New York Solomon KH (1985) Water-salinity-production functions. Transac- Istok JD, Smith JD, Flint AL (1993) Multivariate geostatistical tion: Am Soc Agr Eng 28: 197.5–19 analysis of groundwater contamination: a case history. Ground USSL (1954) Diagnosis and improvement of saline and alkali soils: Water 31:63–74 U.S. Dept. Agr. Handb. 60, 160 p Jang CS, Liou YT, Liang CP (2010) Probabilistically determining Van Meirvenne M, Goovaerts P (2001) Evaluating the probability of roles of groundwater used in aquacultural fishponds. J Hydrol exceeding a site-specific soil cadmium contamination threshold. 388:491–500 Geoderma 102:75–100 Journel AG, Huijbregts CJ (1978) Mining geostatistics. London Academic Press, New York 600 p Juang KW, Lee DY (2000) Comparison of three nonparametric kriging methods for delineating heavy metal contaminated soils. J Environ Qual 29:197–205

123